Underwater Acoustic Signal Processing and Feature Extraction Techniques

Underwater acoustic signal processing encompasses a suite of methods designed to detect, denoise, characterise and classify sound waves propagating through the marine environment. Key challenges stem from multipath propagation, variable ambient noise and non-stationary signal characteristics. Early approaches relied on linear spectral analysis such as the short-time Fourier transform and wavelet decomposition, but these often struggle with overlapping spectral components and transient events. More recent data-driven and adaptive techniques, notably empirical mode decomposition (EMD) and its variants, offer the ability to decompose complex signals into intrinsic mode functions that capture time-varying frequency content. Complementary measures of signal complexity—permutation entropy, dispersion entropy and related metrics—quantify irregularity and dynamical structure, facilitating robust feature extraction. Machine-learning classifiers, particularly support vector machines and neural networks, then map these features onto detection or identification tasks. Together, these advances support applications ranging from passive sonar vessel classification and environmental monitoring of marine mammals to oceanographic sensing and naval surveillance.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Underwater Acoustic Signal Processing and Feature Extraction Techniques publication trend

The graph below shows the total number of articles in underwater acoustic signal processing and feature extraction techniques across all publications each year (not limited to Nature Index journals).

Technical terms

Empirical Mode Decomposition (EMD): A data-driven algorithm that decomposes a non-stationary signal into intrinsic mode functions without predefined basis functions.

Variational Mode Decomposition (VMD): An adaptive decomposition technique that separates a signal into modes by solving a constrained variational problem, improving robustness and mode separation.

Complete Ensemble EMD with Adaptive Noise (CEEMDAN): An enhancement of EMD that adds and subtracts noise adaptively to reduce mode mixing and ensure reliable mode extraction.

Hilbert–Huang Transform (HHT): A two-step approach combining EMD and the Hilbert spectral analysis to produce a high-resolution time-frequency representation of a signal.

Permutation Entropy: A complexity measure that quantifies the randomness and temporal ordering of values in a time series, sensitive to dynamical changes.

Support Vector Machine (SVM): A supervised machine-learning classifier that finds the optimal hyperplane separating feature vectors of different classes with maximum margin.

References

1. HilbertHuang Transform With Intelligent Noise Reduction for Passive SONAR Signal Processing. IEEE Journal of Oceanic Engineering (2025).
2. A New Underwater Acoustic Signal Denoising Technique Based on CEEMDAN, Mutual Information, Permutation Entropy, and Wavelet Threshold Denoising. Entropy (2018).
3. A Novel Feature Extraction Method for Ship-Radiated Noise Based on Variational Mode Decomposition and Multi-Scale Permutation Entropy. Entropy (2017).